JP3918510B2 - Moving picture coding apparatus and moving picture coding method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a moving image encoding apparatus. as well as Video coding Method In particular, in order to efficiently transmit, store, and display images, high-efficiency coding that converts image information into a digital signal with a smaller code amount, three types of in-picture independent, unidirectional prediction, and bidirectional prediction. Video encoding apparatus that facilitates editing of a code string while suppressing a decrease in encoding efficiency in video encoding using the above-described encoding method as well as Video coding Method About.
[0002]
[Prior art]
Conventionally, an MPEG (Moving Picture Experts Group) method is known as a high-efficiency compression encoding method for moving images. This MPEG system has three picture types according to the inter-picture prediction method. They are an intra-picture independent coded picture called an I picture, a unidirectional predictive coded (interframe or interfield forward predictive coded) picture called a P picture, and a bidirectional predictive coded picture called a B picture. The I picture corresponds to random access and channel switching, and can be decoded therefrom. Here, a picture indicates one frame or one field of a moving image.
[0003]
A plurality of pictures are bundled in the code string, and a code string group, that is, a GOP (Group Of Picture) is formed. In this GOP, one I picture is always included. A normal GOP configuration starts with a B picture and ends with a P picture. This picture configuration is shown in FIG. Since the order of the P (I) picture and the B picture is switched in the code string, such a configuration is obtained.
[0004]
On the other hand, since the first B picture is also predicted from the P picture of the previous GOP, inter-picture prediction is not interrupted and the code string cannot be replaced in units of GOP. Therefore, there is a method in which the first B picture is eliminated and the I picture immediately after the P picture of the previous GOP is used. This is called a closed GOP. Since each GOP is not related to the preceding and following GOPs, the code string can be edited in units of GOPs. This GOP configuration is shown in FIG. In this case, since the first part is not a periodic process, the process is somewhat troublesome. In addition, since the B picture with a small code amount is deleted, the average code amount increases.
[0005]
On the other hand, as disclosed in Japanese Patent Application Laid-Open No. 11-164307, the present inventor previously encoded a moving picture coding that multiplexes a code string that is independently encoded in a picture as a sub code string separately from the main code string Apparatus and moving picture decoding apparatus. In this moving image encoding apparatus, main encoding that performs intra-frame independent encoding or inter-image predictive encoding on an input moving image by switching in units of frames or fields and outputs the obtained main code sequence Sub-encoding means for independently encoding a predetermined frame or field among frames or fields for which inter-picture predictive encoding is performed in the main encoding means, and outputting an obtained sub-code sequence; It is characterized by comprising code sequence multiplexing means for inserting a sub code sequence of the predetermined frame or field into an adjacent portion of the main code sequence of the predetermined frame or field to obtain a multiplexed code sequence.
[0006]
In addition, the moving image decoding apparatus detects a type of a code string (main code string / sub code string) to be input from a header of the code string and outputs type information of the code string; Based on the type information of the code string, when the continuous image is not decoded, any code string to be input is led to the decoding process, and when the continuous image is decoded, A code sequence control unit that abandons the sub code sequence and guides only the main code sequence to the decoding process, and a code sequence provided from the code sequence control unit, performs intra-picture decoding or inter-picture predictive decoding, And a decoding unit that outputs the obtained reproduced image.
[0007]
[Figure 1]
According to the moving picture coding apparatus and moving picture decoding apparatus proposed previously by the present inventor, in normal decoding, decoding is performed without using an intra-picture independent coded code string, and random access or channel Only at the time of switching, decoding can be performed from a code string that has been independently encoded within a picture.
[0008]
In addition, there is a technique for improving image quality using both a locally decoded image obtained by independent intra-picture coding and an inter-picture prediction image. For example, in the moving picture coding apparatus previously disclosed in Japanese Patent Laid-Open No. 5-130591, the present inventor adaptively adds a reproduction picture of intra-picture independent coding and an inter-picture prediction picture to form a prediction signal. To do.
[0009]
FIG. 7 is a block diagram showing an example of a conventional moving picture encoding apparatus. In the figure, all the moving image signals coming from the image input terminal 1 are supplied to the frame delay device 2, while only the signal to be encoded as an I picture is supplied to the DCT 20 via the switch 19. The frame delay unit 2 delays only the B picture by a frame time in order to encode the P picture before the B picture. Each image whose order has been changed is given to the subtracter 3.
[0010]
The image signal from the frame delay unit 2 is subtracted from a prediction signal from an adder 9 (to be described later) in a subtracter 3 to obtain a prediction residual, which is input to the DCT 4. The DCT 4 performs a discrete cosine transform (DCT) conversion process on the prediction residual and supplies the obtained coefficient to the quantizer 5. The quantizer 5 quantizes the input coefficient with a predetermined step width, and supplies the coefficient that has become a fixed-length code to the variable-length encoder 6 and the inverse quantizer 10. The variable-length encoder 6 compresses the fixed-length prediction residual with the variable-length code and supplies the obtained code to the multiplexer 13.
[0011]
On the other hand, the inverse quantizer 10 and the inverse DCT 11 perform the inverse processing of the DCT 4 and the quantizer 5 to reproduce the prediction residual. The obtained reproduction prediction residual is added by the adder 12 with the prediction signal from the adder 9 to form a reproduced image, which is input to the inter-picture predictor 7. The inter-picture predictor 7 forms an inter-picture prediction signal using this reproduced picture as a reference picture and supplies it to the multiplier 8. The multiplier 8 multiplies the reproduced image by a value from 0 to 1 in accordance with control information from the specific image setting unit 18 described later, and supplies the multiplied image to the adder 9.
[0012]
The coding of the I picture is performed for a part of the images set periodically among the images coded as the P picture. The encoding of the I picture is similar to the above processing for the prediction residual, and is encoded by the circuit unit including the DCT 20, the quantizer 21, and the variable length encoder 22. This processing is performed by the I (P) picture. This is the same as the processing of the circuit unit comprising the DCT 4, the quantizer 5 and the variable length encoder 6. The obtained code is input from the variable length encoder 22 to the multiplexer 13.
[0013]
On the other hand, in the inverse quantizer 15 and the inverse DCT 16, the inverse processing of the DCT 20 and the quantizer 21 is performed to reproduce an image. The obtained reproduced image (I picture local decoded image) is supplied to the multiplier 17. The multiplier 17 multiplies the locally decoded image by a value from 0 to 1 according to control information from the specific image setting unit 18 described later, and supplies the result to the adder 9.
[0014]
The adder 9 adds the inter-picture prediction image from the multiplier 8 and the I picture local decoded image from the multiplier 17 to obtain a final prediction image. The multiplication coefficient of the multiplier 8 and the multiplication coefficient of the multiplier 17 have a sum of 1 and may be controlled by correlation of images. For a non-specific picture without an I picture, 1 is multiplied by the multiplier 8 and 0 is multiplied by the multiplier 17, and normal P picture processing is performed. Since the B picture does not become a prediction reference image, this addition processing is irrelevant.
[0015]
The specific image setting unit 18 sets a P picture for each predetermined period as a specific picture, and supplies the control information to the switch 19, multipliers 8 and 17, and the multiplexer 13. The multiplexer 13 multiplexes the information of the specific picture and the code string of each picture, and outputs them from the code string output terminal 14.
[0016]
Next, a conventional moving image code string will be described. The code string configuration of a conventional GOP (image group) is as shown in FIG. 8A for a normal GOP and as shown in FIG. 8B for a closed GOP. In FIG. 8, a delimiter indicates a code string of each picture, I, B, and P are picture types, and numbers are reproduction display picture numbers. In the code string, it can be seen that the order of the B picture and the P (I) picture is reversed. As a result, the last GOP is not a P picture but a previous B picture.
[0017]
When a moving image code string having a normal GOP configuration is edited in units of GOP, the first B picture cannot be decoded. This is because the previous P picture belongs to the previous GOP, and the P picture is changed to another image by editing in GOP units, so that a correct reference image cannot be obtained. In this case, in order to prevent the decoding apparatus from decoding the B picture image, it is necessary to set a flag (Bloken Link) indicating that editing is being performed.
[0018]
On the other hand, a moving image code string having a closed GOP configuration does not have the first B picture, so that even if editing is performed in units of GOP, decoding is not affected. This is because the inter-picture prediction is closed by GOP, and a flag (Bloken indicating that editing is being performed)
Link) is not necessary.
[0019]
The conventional moving picture decoding apparatus corresponding to the conventional moving picture encoding apparatus shown in FIG. 7 is adapted to use the independent frame decoded image and the inter-picture prediction image in the formation of the prediction signal as in the local decoding portion of FIG. It is the structure which adds automatically.
[0020]
On the other hand, when a video code string having a normal GOP structure is edited in units of GOPs and a Broken Link flag is set, in decoding, B pictures before the I picture after the editing point are Replace with previous image etc. without decoding. A moving image code string having a closed GOP configuration is not affected by code string editing in units of GOPs, but the period of the P picture is discontinuous, so that a decoding process corresponding thereto is required.
[0021]
[Problems to be solved by the invention]
In the conventional video editing, a code sequence group bundled with a cycle of I pictures, that is, a code sequence having a GOP (Group Of Picture) unit, is edited. However, it is difficult to edit the code sequence in a normal GOP configuration. In the closed GOP configuration without the first B picture, there are problems that the encoding efficiency is lowered and the period of the P picture becomes discontinuous.
[0022]
In addition, the conventional method having an I picture as a subcode sequence in a P picture is effective for random access or the like, but the code amount increases by the number of overlapping I pictures, resulting in a GOP structure corresponding to code sequence editing. Not.
[0023]
Furthermore, the conventional method of forming a prediction signal from an I picture locally decoded image and an inter-picture prediction signal of the same frame has good encoding efficiency, but cannot be decoded without both of the code strings. I can't.
[0024]
The present invention has been made in view of the above points. The predetermined P picture also has an I picture, and the sum of the two reproduced images is used as a reproduced image, so that the quality of the reproduced image can be improved while being editable. Video encoding device as well as Video coding Method The purpose is to provide.
[0025]
[Means for Solving the Problems]
In order to achieve the above object, the moving picture coding apparatus according to the present invention is a moving picture coding that encodes each picture of a moving picture by three kinds of coding methods of intra-picture independent, unidirectional prediction, and bidirectional prediction. In the apparatus, a first encoded local decoding unit that encodes an input image signal by unidirectional prediction or bi-directional prediction and locally decodes to obtain a local decoded image of unidirectional predictive encoding, and encodes by unidirectional prediction Second coded local decoding in which a part of the picture to be processed is defined as a specific picture, and the specific picture is also encoded with independent decoding within the picture together with unidirectional prediction coding, and is locally decoded to obtain a locally decoded image of independent coding within the picture And inter-picture prediction means for adding a locally decoded image of unidirectional predictive coding and a locally decoded image of intra-picture independent coding to obtain a reference image for inter-picture prediction processing of other pictures in a specific picture. It is obtained by a configuration in which.
[0026]
In the present invention, in addition to a picture code string that is one-way predictive coded in a specific picture, it also has a picture code string that is independently coded within a picture, and these two types of picture code strings overlap. The S / N of the reproduced image can be improved by adding the local decoded images, and the inter-image prediction efficiency can be improved by using the reproduced image as a reference image for inter-image prediction.
[0027]
In order to achieve the above object, the moving picture coding method of the present invention is a moving picture in which each picture of a moving picture is coded by three kinds of coding methods of intra-picture independent, unidirectional prediction, and bidirectional prediction. In the encoding method, a first step of encoding an input image signal by unidirectional prediction and local decoding to obtain a locally decoded image of unidirectional prediction encoding; and a part of a picture encoded by unidirectional prediction A second step in which a specific picture is encoded, and the specific picture is also encoded in the picture independently and is independently decoded, and is locally decoded to obtain a locally decoded image of the intra-picture independent encoding; And a third step of adding the locally decoded image and the locally decoded image of independent intra-picture coding to obtain a reference image for inter-picture prediction processing of other pictures.
[0028]
In the present invention, in addition to a picture code string that is one-way predictive coded in a specific picture, it also has a picture code string that is independently coded within a picture, and these two types of picture code strings overlap. The S / N of the reproduced image can be improved by using the signal obtained by adding the locally decoded images as a reference image for inter-picture prediction processing of other pictures, and the reproduced image can be used as a reference image for inter-picture prediction. Thus, the inter-picture prediction efficiency can be improved.
[0031]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a moving picture encoding apparatus according to the present invention. In the figure, the same components as in FIG. In this specification, “picture” refers to one frame or field.
[0032]
In FIG. 1, all the moving image signals coming from the image input terminal 1 are given to the frame delay unit 2, and only what is encoded as an I picture is given to the DCT 20 via the switch 19. The frame delay unit 2 delays only the B picture in order to encode the P picture before the B picture. Each image whose order has been changed is given to the subtracter 3.
[0033]
The input image signal delayed by the frame delay unit 2 is subtracted from the prediction signal supplied from the inter-picture predictor 27 in the subtracter 3 to obtain a prediction residual, which is input to the DCT 4. The DCT 4 performs DCT (Discrete Cosine Transform) conversion processing on the prediction residual, and gives the obtained coefficient to the quantizer 5. The quantizer 5 quantizes the given coefficient with a predetermined step width, and supplies the coefficient having a fixed length code to the variable length encoder 6 and the inverse quantizer 10. The variable length encoder 6 compresses the fixed length prediction residual from the quantizer 5 with the variable length code, and the obtained variable length code of the P picture or B picture is supplied to the multiplexer 13.
[0034]
On the other hand, the inverse quantizer 10 and the inverse DCT 11 perform the inverse processing of the DCT 4 and the quantizer 5 to reproduce the prediction residual. The obtained reproduction prediction residual is added to the prediction signal from the inter-picture predictor 27 in the adder 12 to form a locally decoded image, which is supplied to the multiplier 25. The multiplier 25 multiplies the locally decoded image by a value from 0 to 1 according to the control information from the specific image setting unit 18 and supplies the result to the adder 26.
[0035]
The coding of the I picture is performed for a part of the images set periodically among the images coded as the P picture. The encoding of the I picture is performed in the same manner as the above processing for the prediction residual. That is, the I picture is input to the variable length encoder 22 through the DCT 20 and the quantizer 21 and is variable length encoded. This process is performed by the DCT 4, the quantizer 5 and the variable length code for the P (B) picture. This is the same as the processing of the generator 6. The variable length code of the I picture obtained by the variable length encoder 22 is input to the multiplexer 13.
[0036]
On the other hand, the inverse quantizer 15 and the inverse DCT 16 perform the inverse processing of the DCT 20 and the quantizer 21 to reproduce the locally decoded image. The obtained locally decoded image is given to the multiplier 17. The multiplier 17 multiplies the local decoded image by a value from 0 to 1 according to the control information from the specific image setting unit 18 and supplies the result to the adder 26.
[0037]
The adder 26 adds the two types of locally decoded images from the multipliers 25 and 17 to obtain a reference image for inter-picture prediction processing. The inter-picture predictor 27 forms an inter-picture prediction signal using this reference picture. This inter-picture prediction signal is supplied to the subtracter 3 and the adder 12, respectively.
[0038]
The specific image setting unit 18 sets a P picture for each predetermined period as a specific picture, and supplies the control information to the switch 19, multipliers 17 and 25, and the multiplexer 13. The multiplexer 13 multiplexes the information of the specific picture and the code string of each picture, and outputs them from the code string output terminal 14. The switch 19 is turned on only for the above-mentioned specific picture, and is turned off for the other non-specific pictures.
[0039]
Next, addition processing of two types of locally decoded images in the adder 26 will be described. First, since there is no I picture in a non-specific picture, the multiplier 25 multiplies 1 and the multiplier 17 multiplies 0. That is, it is not different from general P picture encoding. In addition, since the B picture does not become a reference image, the addition process is not related in the first place.
[0040]
On the other hand, in the specific picture, both the multiplier 25 and the multiplier 17 multiply the input local decoded image by the coefficient 0.5 in order to add the locally decoded image of the P picture and the locally decoded image of the I picture. When the noise component included in each image is white noise, the S / N of 3 dB can be improved by addition, but the noise component of the locally decoded image of the P picture and the noise component of the locally decoded image of the I picture are processed separately. Although the method is different, there are some common points such as coarse quantization with high frequency components, so there is also a correlation with noise components, and an improvement of 3 dB cannot be obtained. However, since they are not the same, some improvement is expected. Assuming that the half is 1.5 dB, it is necessary to increase the code amount by about 30% in order to improve the code amount accordingly.
[0041]
Generally, by setting the quantization step widths of the quantizers 5 and 21, the I picture is set to have a higher quality of the reproduced image than the P picture. This is because increasing the quality of the I picture that is the basis of the reference image of all images in the GOP contributes to improving the image quality of the entire GOP. On the other hand, if the S / N is different between the reproduced picture of the P picture and the reproduced picture of the I picture, the addition is not very effective. Therefore, if the code amount of the I picture is reduced to some extent, the P picture and the S / N become equivalent, and the maximum effect can be obtained.
[0042]
In the present invention, since the P picture is added to the normal GOP configuration, the code amount increases accordingly. However, even if the code amount of the I picture is reduced and the S / N is reduced, the I picture and the P picture are reduced. If the S / N of the reference image can be maintained by addition, the reproduced image and the code amount are equivalent to those of a normal GOP.
[0043]
Here, the generated code amount will be compared with a normal GOP and a closed GOP. Assume that the average code amount of an I picture is 1000 kbit, the average code amount of a P picture is 300 kbit, and the average code amount of a B picture is 100 kbit. In the case of a normal GOP with an image of 30 frames per second and a P (I) picture cycle of 3 frames, if the GOP length is 15 frames, the average transfer rate is 6.4 Mbps from the average number of each picture per second. It becomes.
[0044]
On the other hand, in the case of a closed GOP, the size of the GOP is different from that of the normal GOP, the GOP length is 13 frames, the average transfer rate is 6.92 Mbps, the GOP length is 16 frames, and the average transfer rate is The average transfer rate increases to 6.56 Mbps compared to the normal GOP. In addition, the accessibility is slightly improved when the GOP length is 13 frames, but it is lowered when the GOP length is 16 frames. If a code amount corresponding to 15 frames is obtained from both, it is 6.68 Mbps, which is a 4.4% increase in code amount with respect to a normal GOP.
[0045]
In the case of this embodiment, if the average code amount of an I picture is the same as that of a normal GOP or closed GOP, the average transfer rate is 7.0 Mbps, but if it is reduced by 30% to 700 kbit, the average transfer rate is 6.4 Mbps. This is the same as in the case of normal GOP. This is a form in which the code amount of an I picture of a normal GOP is allocated to an I picture and a P picture.
[0046]
Next, the moving image code string will be described. In forming a code string encoded by the encoding apparatus shown in FIG. 1, the P picture code string of a specific picture is the last of the GOP (image group), and the I picture is the first of the GOP. Therefore, in a specific picture, a code string is arranged in the order of P picture and I picture, and when viewed in one GOP, it starts with I picture and ends with P picture. The GOP configuration of this embodiment is shown in FIG.
[0047]
On the other hand, since the B picture and the P (I) picture are reversed in the code string, the last B picture is not the P picture but the previous B picture. That is, as shown in FIG. 8C, the code sequence of the GOP (image group) to be formed starts with an I picture code sequence I1 that is independently encoded within a picture of a specific picture, and immediately before the next specific picture. The process ends with a B picture code string B15 that has been bi-predictively encoded.
[0048]
This GOP configuration is similar to a closed GOP when only one GOP is compared while ignoring the overlap of specific frames. When one of the I picture or P picture is deleted in a specific frame, the GOP configuration is determined by the deleted one. Changes, but the picture sequence is usually the same as the GOP sequence. That is, the GOP according to the present embodiment can have both the characteristics of a closed GOP and a normal GOP.
[0049]
The moving image code string can be edited in units of GOP as in the case of a closed GOP. This is shown in FIG. As shown in the drawing, 1 GOP of the code string B shown in the third line is inserted between the GOP and GOP having the code string A shown in the first line, and the edited code as shown in the second line A column is obtained. Here, each GOP has overlapping pictures at the beginning and end. Therefore, the processing is different from the conventional editing apparatus.
[0050]
First, regarding the image length, the last P picture of the GOP is not included in the GOP length (time), and the editing time is calculated. Therefore, even if the code sequence of the GOP configuration of this embodiment is 16 frames, it is regarded as 15 frames.
[0051]
Next, when editing is performed in units of GOPs by reproduction control of a specific picture, both the P picture of the previous GOP and the I picture of the subsequent GOP can be decoded and reproduced as the specific picture serving as the editing point. On the other hand, since editing is performed, the image contents are different. As a method of positively using the point where the code strings overlap, if the control information is entered as to which image is output at the time of reproduction, the edit point can be moved back and forth by one picture with the same code string.
[0052]
Further, in the conventional closed GOP, since there is no processing change in the decoding device, it is not necessary to set a flag (Bloken Link) indicating that editing is performed, but in this method, the decoding processing is switched. Since it is necessary, it is necessary to set a flag of Broken Link.
[0053]
Next, examples of the video decoding device will be described. FIG. 3 is a block diagram illustrating an example of a moving picture decoding apparatus. This moving picture decoding apparatus shows a configuration of a decoding apparatus corresponding to the embodiment of the moving picture encoding apparatus of the present invention shown in FIG. 1, and this is an image continuity without editing. This is a case of a code string in which is maintained.
[0054]
In FIG. 3, a code string coming from a code string input terminal 31 is separated into a code string of I picture and other code strings by a demultiplexer 32 based on the header of the picture. The code sequence of P picture and B picture is supplied to the variable length decoder 33, and the code sequence of I picture is supplied to the variable length decoder 34.
[0055]
The code sequence of the P (B) picture is returned to the fixed length code by the variable length decoder 33, and the variable length code of the prediction residual is supplied to the inverse quantizer 35. The inverse quantizer 35 inversely quantizes the input fixed length code according to the quantization parameter to obtain a reproduction DCT coefficient value of the prediction residual, and supplies this to the inverse DCT 36.
[0056]
The inverse DCT 36 converts 8 × 8 coefficients into a decoded prediction residual signal and supplies it to the adder 37. The adder 37 adds the prediction signal given from the inter-picture predictor 45 to the decoded prediction residual signal to obtain a decoded image signal. The decoded image signal of the P (B) picture obtained in this way is supplied to the multiplier 42.
[0057]
On the other hand, the I-picture code string separated by the demultiplexer 32 is decoded by the variable length decoder 34, dequantized by the inverse quantizer 38, decoded by the inverse DCT 39, and reproduced image signal. Is input to the multiplier 41. The operations of the variable length decoder 34, the inverse quantizer 38, and the inverse DCT 39 are the same as those of the variable length decoder 33, the inverse quantizer 35, and the inverse DCT 36, but the parameters are for I picture.
[0058]
The demultiplexer 32 detects the picture ID from the picture header in the input code string, and the result information is supplied to the specific image controller 40. The specific image controller 40 detects the specific picture and supplies the control information to the multipliers 41 and 42, respectively. The multiplier 41 multiplies the reproduced image signal from the inverse DCT 39 by a value from 0 to 1 in accordance with the control information described above, and gives the result to the adder 43. On the other hand, the multiplier 42 multiplies the decoded image signal of the P (B) picture from the adder 37 by a value from 0 to 1 according to the control information described above, and supplies the result to the adder 43.
[0059]
The adder 43 adds the two types of decoded image signals extracted from the multipliers 41 and 42 to obtain a reproduced image signal. Addition by the adder 43 is performed only for a specific picture. At this time, both the multiplier 41 and the multiplier 42 are multiplied by a coefficient 0.5. Otherwise, the multiplier 42 multiplies the coefficient 1 and the decoded image signal, and the multiplier 41 multiplies the coefficient 0 and the decoded image signal. B) The decoded image signal of the picture is output as it is.
[0060]
In a specific picture, the reproduced image signal is improved by S / N from the decoded image signals extracted from the multipliers 41 and 42 by addition in the adder 43. A state of such decoding is shown in FIG.
[0061]
The reproduced image signal output from the adder 43 is output as it is from the reproduced image output terminal 47 via the switch 46 in the B picture. On the other hand, the reproduced image signal output from the adder 43 is temporarily stored in the image memory 44 in the P (I) picture, is used as a reference image for inter-image prediction processing, and is delayed. The reference image is used as a prediction signal and input to the adder 37. The switch 46 selects the delayed B picture and the P (I) picture delayed in the image memory 44 and outputs the selected picture to the output terminal 47.
[0062]
Next, another example of the moving picture decoding apparatus will be described. FIG. 5 shows a block diagram of another example of a moving picture decoding apparatus corresponding to the moving picture encoding apparatus of FIG. 5, the same components as those in FIG. 3 are denoted by the same reference numerals, and description thereof is omitted. The decoding apparatus of FIG. 5 is an apparatus that performs decoding when code sequence editing is performed and image continuity is not maintained.
[0063]
The decoded picture signal of the B picture output from the adder 37 is output as it is from the reproduced picture output terminal 47 via the switch 51. On the other hand, the decoded picture signal of the P picture output from the adder 37 is temporarily held in the picture memory 48. Also, the decoded picture signal of the I picture output from the inverse DCT 39 is temporarily held in the picture memory 49.
[0064]
Here, since the GOP has been edited, the P picture and I picture of the specific picture are formally the same picture, but the P picture is that of the previous GOP, and the I picture is that of the subsequent GOP. Is. Therefore, the switch 50 selects an appropriate one as a reference image from the two types of decoded image signals from the image memories 48 and 49 as follows.
[0065]
Following the decoding of the specific picture, the B picture of the previous GOP is decoded by selecting the decoded picture of the P picture held in the picture memory 48. Subsequently, in decoding of the P picture and B picture of the next GOP, a decoded image of the I picture held in the image memory 49 is selected. The state of decoding in this case is shown in FIG. In the figure, the arrows indicate the inter-image prediction relationship.
[0066]
In the decoding of the decoding apparatus shown in FIG. 5, the reference image for inter-picture prediction is slightly different from the reference image of the encoding apparatus, but both are decoded images for the same image, except for the quantization noise component. The original image is common. The change in the reference image affects only 2 pictures immediately before the editing point and 1 GOP after the editing point. However, since the prediction residual components are sequentially added after the editing point, the influence of the reference image change gradually decreases. On the other hand, considering the visual characteristics, it is said that when the scene changes during editing, it is difficult to notice deterioration, and the detection capability is said to be greatly reduced for about 0.1 seconds immediately after the change. Therefore, the visual impact of degradation is very small.
[0067]
The playback image output is selected by the switch 51. The operation of the switch 51 is the same as that of the switch 46 in FIG. 3 except for the specific picture. Since it is possible to output either an I picture or a P picture in a specific frame, it may be determined in advance which one is selected, but the control information is put in the code string by the code string editing device. If so, control accordingly.
[0068]
In the above embodiment, the moving picture encoding apparatus and method have been described. However, the present invention is not limited to this, and a part of a picture encoded by unidirectional prediction is a specific picture. In the specific picture, there is a code sequence that is encoded independently within the picture as well as one-way predictive encoding, and starts with a code sequence that is independently encoded within the picture of the specific picture, and bidirectional predictive encoding of the picture immediately before the specific picture is performed The code string ending with the code string may be defined as one code string group, and the code string synthesized in units of the code string group may be transmitted via a desired transmission path.
[0069]
【The invention's effect】
As described above, according to the present invention, a specific picture has a picture code string that is independently encoded within a picture in addition to a picture code string that is unidirectionally predictively encoded, and these two types of picture code strings are Although overlapping, the S / N of the reproduced image is improved by adding the locally decoded images of both, so that the quality of the reproduced image can be improved. In addition, since the inter-picture prediction efficiency is improved by using the reproduced picture as a reference picture for inter-picture prediction, the total amount of codes can be reduced correspondingly in the specific picture.
[0070]
Further, according to the present invention, the code sequence group (GOP) configuration is divided between the unidirectional predictive coded picture included in the specific picture and the intra-picture independent coded picture, so that the code is encoded in GOP units. At a discontinuous point when a column is edited, the decoding apparatus performs inter-picture prediction of another picture using a unidirectional predictive coded picture as a GOP end and an intra-picture independent coded picture as a reference picture at the GOP end. Therefore, it is possible to generate a code string that can be edited in units of GOPs in the same manner as a closed GOP while maintaining the periodicity of a unidirectional predictive coded picture with a coding efficiency equivalent to that of a normal GOP.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of a moving image encoding apparatus of the present invention.
FIG. 2 is a diagram illustrating an example of how a moving image code string is edited.
FIG. 3 is a block diagram illustrating an example of a moving picture decoding apparatus.
FIG. 4 is a diagram illustrating each example of decoding in units of pictures.
FIG. 5 is a block diagram of another example of the video decoding device.
FIG. 6 is a diagram illustrating each example of a GOP configuration.
FIG. 7 is a block diagram of an example of a conventional video encoding device.
FIG. 8 is a diagram illustrating each example of a code string configuration of a GOP (image group).
[Explanation of symbols]
1 Image input terminal
2 frame delay
3 Subtractor
4, 20 DCT
5, 21 Quantizer
6, 22 Variable length encoder
10, 15, 35, 38 Inverse quantizer
11, 16, 36, 39 Inverse DCT
12, 26, 37, 43 Adder
13 Multiplexer
14 Code string output terminal
17, 25, 41, 42 multiplier
18 Specific image setting device
19, 46 switch
27, 45 Image predictor
31 Code string input terminal
32 Demultiplexer
33, 34 Variable length decoder
40 Specific image controller
47 Playback image output terminal

Claims (2)

In a moving image encoding apparatus that encodes each picture of a moving image with three types of encoding methods of independent in picture, unidirectional prediction, and bidirectional prediction,
A first encoded local decoding unit that encodes an input image signal by the unidirectional prediction or bidirectional prediction, and locally decodes to obtain a locally decoded image of unidirectional predictive encoding;
A part of a picture to be encoded by the unidirectional prediction is a specific picture, and the specific picture is encoded by unidirectional predictive encoding and independent in the picture, is locally decoded, and is locally decoded. Second encoded local decoding means for obtaining
Inter-picture prediction means for adding the locally decoded image of the one-way predictive coding and the locally decoded image of the intra-picture independent coding to the reference picture of the inter-picture prediction processing of other pictures in the specific picture. A moving picture coding apparatus characterized by the above. In the moving picture coding method for coding each picture of a moving picture by three kinds of coding methods of independent in picture, unidirectional prediction, and bidirectional prediction,
A first step of encoding an input image signal by the unidirectional prediction and locally decoding to obtain a locally decoded image of unidirectional predictive encoding;
A part of a picture to be encoded by the unidirectional prediction is a specific picture, and the specific picture is encoded by the unidirectional predictive encoding as well as the intra-picture independent, and is locally decoded to generate a locally decoded image of the intra-picture independent encoding. A second step of obtaining
A third step of adding the locally decoded image of the one-way predictive coding and the locally decoded image of the intra-picture independent coding to the reference picture of the inter-picture prediction process of another picture in the specific picture. A video encoding method characterized by the above.

Images